MEG Investigations of Spectral and Temporal Resolution Properties of Human Auditory Cortex
CNL
MEG
Behavioral
Psychophysics
Genetics
Nicole M. Gage, PhDDepartment of Cognitive Sciences
University of California, Irvine
•Neurobiology of language dysfunction in developmental disorder
cognitiveneuroscienceof languageLaboratory
CNL
•Speech perception and hemispheric asymmetries in speech processing
•Cortical language function and mapping in healthy adults and pre-surgical patients
MEGBehavioralPsychophysicsGenetics
•Auditory perception and cortical sound processing
•History of neuroscience: Carl Wernicke’s model for language and his theory of conceptual representation in cortex
cognitiveneuroscienceof languageLaboratory
CNL
Research Goals:
To understand neural mechanisms that underlie speech and language processing in healthy adults and typically developing children
To elucidate neural processes that underlie language dysfunction in developmental disorder, such as autism
To understand the correspondence between genetics, brain, and behavior in the language domain
MEGBehavioralPsychophysicsGenetics
•Neurobiology of language dysfunction in developmental disorder
cognitiveneuroscienceof languageLaboratory
CNL
•Speech perception and hemispheric asymmetries in speech processing
•Cortical language function and mapping in healthy adults and pre-surgical patients
MEGBehavioralPsychophysicsGenetics
•Auditory perception and cortical sound processing
•History of neuroscience: Carl Wernicke’s model for language and his theory of conceptual representation in cortex
Plan of the Talk
Studies of cortical sound processing in adultsSpectral resolution properties of auditory cortexIntegrative processes underlying cortical evoked componentsTemporal resolution properties of auditory cortex
Cortical sound processing in typically developing children and children with autism
Spectral resolution for speech and non-speech soundsMaturational changes in cortical evoked componentsTemporal resolution properties of auditory cortex
A case study: Child with autism and language impairment, a rare chromosome deletion on a region implicated in language, and extreme sensory reactivity
Temporal Resolution of Auditory Cortical Systems
The temporal resolution of the auditory system is exquisite, with neural systems that decode features in the acoustic signal capable of submillisecond resolution.
The high level of resolution in auditory cortical systems provides the capability for decoding fine-grained fluctuations in sounds, critical to the accurate perception of speech.
Magnetoencephalography (MEG)
• Millisecond temporal resolution• Post-synaptic, dendritic flow• Synchronized response of populations of neurons• Time-locked to a stimulus event• Modeled by a single equivalent current dipole
Neuromagnetic Auditory Evoked Field
Weak FieldWeak Field
Strong FieldStrong Field
Iso-Field ContoursIso-Field Contours
Recording SurfaceRecording Surface
Magnetic Fields Sources Orientation of Neurons
Right Left
M100Dipole
Detection DeviceLiquid Helium
SQUID
SuperconductingCoils
Magnetic Field
Basic Principles of MEG
Magnetic Field Pattern Model
Nose
Left Right
Sensor coils
148 Channel Sensor Array
Magnetic Field Contour Map
Left and Right Hemisphere Auditory Cortical Dipolar Activity
0 100 200Time (ms)
M100
A prototype auditory evoked neuromagnetic field detected by MEG; 37 channels with y-scale representing evoked response magnitude in units of femtotesla (fT) are shown collapsed on the same horizontal time axis.
Neuromagnetic Auditory Evoked Field
M50
Right Left
M100Dipole
M100 localizes to auditory cortex
Nose
Left Right
M100
Frequency Dependence of the M100: In healthy adults, M100 latency is modulated by tone frequency, with longer latencies for low (100-200 Hz) as compared to high (1000-3000 Hz) frequency tones.
For sinusoidal tones, M100 latency is modulated as a function of tone frequency, with a ‘fixed cost of ~100 ms plus a period dependent time that is roughly equal to 3 periods of the sinusoid (~30 ms for a 100 Hz, ~3 ms for a 1kHz tone). The dynamic range of frequency modulation in adults is ~25 ms.
M100 Latency is modulated by tone frequency: sinusoidal
tones 100-1000 Hz
95
100
105
110
115
120
125
M10
0 La
tenc
y (m
s)
100 300 400 500 600 700 1000
Tone Frequency (Hz)
Tone Continuum Response Latency
Delta 15-30 ms
Vowel Continuum varying in values for F1 but otherwise matched.
/u/ /a/F0 100 Hz 100 HzF1 250 Hz 50 Hz steps 750 HzF2 1000 Hz 1000 HzF3 2500 Hz 2500 Hz
Frequency of F1 is inversely related to vowel height, with lower F1associated with high vowels (/u/) and higher F1 with low vowels (/a/).
Investigative Question:Will M100 latency reflect the spectral center of gravity of 3 formant vowels (curvilinear function) or vowel identity (stepped function)?
M100 Role in Speech PerceptionDoes the M100 reflect sensory (acoustic)
or perceptual (representational) processes?
200250300350400450500550600650700750
/u/ /u/ /u/ ambiguous ambiguous ambiguous /a/ /a/ /a/
Tone Frequency (Hz)
Vowel Continuum
80
90
100
110
120
130
140
M100 Latency Prediction - F1 Frequency
80
90
100
110
120
130
140
M100 Latency Prediction - Vowel Identity
160
170
180
190
200
210
220
M10
0 La
tenc
y (m
s)
Stimulus Class
Vowel Continuum Response Latency
GROUP Mean 214.2 214.2 211.6 201.3 199.9 201.3 196.3 196.0 188.1 188.2 186.0/u/ /u/ /u/ amb amb amb amb amb /a/ /a/ /a/
M100 latency reflects vowel identity as well as secondary spectral features in speech sounds
M100 amplitude reflects experience with speech sounds, with lower response amplitudes to novel tokens.
0.50.60.7
0.80.91.01.1
1.21.3
/u/ /u/ /u/ amb amb amb amb amb /a/ /a/ /a/
Stimulus Class
Vowel Continuum Response Amplitude
Neural mechanisms underlying the M100 component reflect phonetically-relevant features in speech
M100 latency reflects vowel identity as well as secondary spectral features in speech sounds
M100 amplitude reflects experience with speech sounds, with lower response amplitudes to novel speech-like tokens.
160
170
180
190
200
210
220
M10
0 La
tenc
y (m
s)
Stimulus Class
Vowel Continuum Response Latency
GROUP Mean 214.2 214.2 211.6 201.3 199.9 201.3 196.3 196.0 188.1 188.2 186.0/u/ /u/ /u/ amb amb amb amb amb /a/ /a/ /a/
0.50.60.7
0.80.91.01.1
1.21.3
/u/ /u/ /u/ amb amb amb amb amb /a/ /a/ /a/
Stimulus Class
Vowel Continuum Response Amplitude
Roberts, Flagg, & Gage, 2004
Boon
0 10040
Time (ms)
Stimulus Onset
Time (ms)
0 40 100
The M100 component has a brief (~35 ms) and finite integrative window during which stimulus attributes are accumulated in the processes leading to the formation of the M100 peak.
Within this integrative window, it is stimulus presence -- and not peak or integrated energy -- that dominates the processes underlying the M100.
A Temporal Window of Integration for the M100
Gage & Roberts, 2000
M100 is highly sensitive (within a brief integrative window) to transient features in consonants that cue distinctive feature contrasts in speech, such as manner and place of articulation, voicing.
The selective activation of the M100 for some stimulus features (periodicity, formant transitions) and not others (absolute sound level) has led to its description as an intermediate processing stage between sensory (acoustic) and perceptual (representational) processing.
M100 Latency for Within-Speech Contrasts
90
95
100
105
110
115
120
Left Hemisphere Right Hemisphere
Stops No-stops
Gage et al., 1998, Gage et al., 2002
M100 Latency for Place Contrasts
0.95
0.97
0.99
1.01
1.03
Left Hemisphere Right Hemisphere
/ba/ /da/ /ga/
Boon
0 10040
Time (ms)
Stimulus Onset
Time (ms)
0 40 100
What is the Temporal Resolution for Resolving Brief Discontinuities in Sounds within the M100 Integrative Window?
Temporal Resolution of the Auditory M100:Gap Detection Experiments
Psychophysical investigations of auditory perceptual acuity frequently employ gap detection paradigms, where a silent gap is inserted in a tone or noise burst and the minimum detectable gap is measured
Gap detection thresholds correspond to speech perception acuity,indicating that similar or overlapping neural processes are employed both in detecting brief silent gaps and in resolving the fine structure of the speech signal.
The investigation: we know that the M100 is sensitive to the presence of a stimulus within a brief and finite integrative window.
What are the lower limits of the resolution for brief discontinuities – or the absence of a stimulus – within the M100 window of integration?
Gage, Roberts, & Hickok, In Press 2005
20 600
Time (ms)
40
0257
1015203050
Gap Duration
(ms)How sensitive is the M100 to fine-grained temporal discontinuities in sounds?
We address this question by inserting brief gaps of silence at +10 ms post stimulus onset and measuring M100 modulation as a function of gap duration.
In a second condition, we inserted gaps at +40 ms post onset. Here we predicted that M100 would not be modulated by gaps of silence because the gaps were inserted outside the integrative window.
Temporal Resolution of the Auditory M100:Gap Detection Experiments
Temporal Windowof Integration (~25-40 ms)
LH R2 = 0.93
RH R2 = 0.99
80
100
120
140
0 ms 2 ms 5 ms 7 ms 10 ms 15 ms 20 msGap Duration (ms)
Left Hemisphere Right Hemisphere
Results: M100 Latency is modulated by Gap Duration
LH R2 = 0.93
RH R2 = 0.93
70
90
110
130
0 ms 2 ms 5 ms 7 ms 10 ms 15 ms 20 msGap Duration (ms)
Left Hemisphere Right Hemisphere
Results: M100 Amplitude is modulated by Gap Duration
80
100
120
140
0 2 5 7 10 15 20Gap Duration (ms)
Left Hemisphere Right Hemisphere
Results: M100 is not affected when gaps are inserted at +40 ms post onset
Conclusions
The integrative processes underlying M100 formation are highly sensitive to fine-grained discontinuities in sounds.
M100 sensitivity to the shortest gap (2 ms) corresponds to clinical and behavioral measures of auditory acuity, where detection thresholds have been reported for gaps of <5 ms.
These data provide further evidence for a short (~35 ms) and finite window of integration in the accumulation processes leading to the M100 peak.
Fine-grained Temporal Resolution of the M100
A Finite Temporal Window of Integration for the M100
M100 - ~35 ms TWI Secondary Auditory CortexFeature Discrimination Processes
The Time Course of Auditory Cortical ProcessingIntegrative Windows for the M50 and M100 ComponentsReflect Underlying Sensory and Perceptual Mechanisms
Gage, Hickok, & Roberts, 2005
M50 - ~10 ms TWI Primary Auditory CortexDetection, Habituation Mechanisms